Efficiently approximating Markov tree bagging for high-dimensional density estimation thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Efficiently approximating Markov tree bagging for high-dimensional density estimation

Published on Oct 03, 20112379 Views

We consider algorithms for generating Mixtures of Bagged Markov Trees, for density estimation. In problems defined over many variables and when few observations are available, those mixtures generally

Related categories

Chapter list

Efficiently Approximating Markov Tree Bagging for High-Dimensional Density Estimation00:00
The goal of this research is to improve probabilistic reasoning in high-dimensional problems.00:17
Mixtures of trees build on the good properties of Markov trees. (1)02:45
Mixtures of trees build on the good properties of Markov trees. (2)03:25
Bagging is a good variance reduction method.04:59
We developed approximation strategies to accelerate it.06:07
Key idea of approximation strategies07:18
1 : In the inertial approach, Si is based on the previous tree Ti-1.08:02
2 : In the skeleton-based approach, all Si are equal and based on the rst tree.08:51
Edges are tested for independence before inclusion in S.10:09
We evaluated our algorithms on synthetic and more realistic data sets.11:31
The two approaches are working well.12:28
In uence of the parameter in the Skeleton-based approximation 13:23
Starting by the max-likelihood tree is necessary in the inertial method. (1)14:53
Starting by the max-likelihood tree is necessary in the inertial method. (2)15:47
More realistic data sets (by C. Aliferis, A. Statnikov, I. Tsamardinos & al).16:40
Both approximations are better than a maximal-likelihood tree in two experimental cases.18:00
ecmlpkdd2011_schnitzler_efficiently_01_Page_1818:30
In most cases only the skeleton-based approximation is good.18:55
Conclusions20:11